A5087768439- Semantic Web and Ontologies
- Topic Modeling
- Natural Language Processing Techniques
- Advanced Database Systems and Queries
- Distributed systems and fault tolerance
- Service-Oriented Architecture and Web Services
- Speech and dialogue systems
- Advanced Text Analysis Techniques
- Data Management and Algorithms
- Software Engineering Research
- Distributed and Parallel Computing Systems
- Data Mining Algorithms and Applications
- Advanced Data Storage Technologies
- Biomedical Text Mining and Ontologies
- AI-based Problem Solving and Planning
- Data Visualization and Analytics
- Cryptography and Data Security
- Logic, programming, and type systems
- Advanced Graph Neural Networks
- Peer-to-Peer Network Technologies
- Privacy-Preserving Technologies in Data
- Cloud Computing and Resource Management
- Data Quality and Management
- Parallel Computing and Optimization Techniques
- Logic, Reasoning, and Knowledge
IBM Research - Thomas J. Watson Research Center
2023
University of Calgary
2004-2022
IBM (United States)
2012-2021
Rensselaer Polytechnic Institute
2019
The University of Texas at Austin
2002-2015
University of Manitoba
1994-2004
University of Ottawa
1996-2002
University of Alberta
2002
Building a knowledge base for given domain traditionally involves subject matter expert and engineer. One of the goals our research is to eliminate There are at least two ways achieve this goal: train experts write axioms (i.e., turn them into engineers) or create tools that allow users build bases without having axioms. Our strategy through instantiation assembly generic components from small library.In many ways, creating such library like designing an ontology: What most general kinds...
In the winter 2004 issue of AI Magazine, we reported Vulcan Inc.'s first step toward creating a question‐answering system called Digital Aristotle. The goal that was to assess state art in applied knowledge representation and reasoning (KRR) by asking experts represent 70 pages from advanced placement (AP) chemistry syllabus deliver knowledge‐based systems capable answering questions syllabus. This article reports next realizing Aristotle: present design evaluation results for AURA, which...
Semantic relationships among words and phrases are often marked by explicit syntactic or lexical clues that help recognize such in texts. Within complex nominals, however, few overt available. Systems analyze nominals must compensate for the lack of surface with other information. One way is to load system semantics nouns adjectives. This merely shifts problem elsewhere: how do we define build large semantic lexicons? Another find constructions similar a given nominal, which already known....
Project Halo is a multistaged effort, sponsored by Vulcan Inc, aimed at creating Digital Aristotle, an application that will encompass much of the world's scientific knowledge and be capable applying sophisticated problem solving to answer novel questions. envisions two primary roles for Aristotle: as tutor instruct students in sciences interdisciplinary research assistant help scientists their work. As first step towards this goal, we have just completed six-month pilot phase designed...
Despite some successes, the lack of tools to allow subject matter experts directly enter, query, and debug formal domain knowledge in a knowledge-base still remains major obstacle their deployment. Our goal is create such tools, so that trained engineer no longer required mediate interaction. This paper presents our work on entry part this overall capture task, which based several claims: users can construct representations by connecting pre-fabricated, representational components, rather...
As part of the ongoing project, Project Halo, our goal is to build a system capable answering questions posed by novice users formal knowledge base. In current context, base covers selected topics in physics, chemistry, and biology, question set consists AP (advanced high-school) level examination questions. The task challenging because are linguistically complex often incomplete (assume unstated knowledge), do not have prior system's contents. Our solution involves two parts: controlled...
Ontology designers often distinguish Entities (things that are) from Events happen). It is not obvious how this division admits Roles are, but only in the context of things For example, Person might be considered an Entity, while Employee a Role. A remains independent which he participates. Someone by virtue participating Employment Event. The problem to represent new, there little consensus on solution. In paper, we present ontology finds place for as well representation allows related and...
This paper describes a framework that uses the Semantic Web infrastructure to address semantic interoperability between relational database systems in large-scale environments and at multiple levels of granularities. Given system, we describe formal algorithm use Rs meta-data structural constraints construct its OWL ontology while preserving underlying system. The generated is described using conforming set vocabularies defined an on web. Using this guarantee applications web can work with...
Sara Rosenthal, Ken Barker, Zhicheng Liang. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint (EMNLP-IJCNLP). 2019.
Many AI tasks require determining whether two knowledge representations encode the same knowledge. Solving this matching problem is hard because may content but differ substantially in form. Previous approaches to have used either syntactic measures, such as graph edit distance, or semantic determine "distance" between representations. Although outperform ones, previous research has focused primarily on use of taxonomic We show that not enough mismatches go largely unaddressed. In paper, we...
Oren Melamud, Mihaela Bornea, Ken Barker. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint (EMNLP-IJCNLP). 2019.
Data warehousing is one of the major research topics appliedside database investigators. Most work to date has focused on building large centralized systems that are integrated repositories founded pre-existing upon which all corporate-wide data based. Unfortunately, this approach very expensive and tends ignore advantages realized during past decade in area distribution support for localization a geographically dispersed corporate structure. This investigates distributed warehouses with...
In this paper, we present the event detection models and systems have developed for Multilingual Protest News Detection - Shared Task 1 at CASE 2021. The shared task has 4 subtasks which cover different granularity levels (from document level to token level) across multiple languages (English, Hindi, Portuguese Spanish). To handle data from languages, use a multilingual transformer-based language model (XLM-R) as input text encoder. We apply variety of techniques build several that perform...
Making logic-based AI representations accessible to ordinary users has been an ongoing challenge for the successful deployment of knowledge bases. Past work meet this objective resulted in a variety ontology editing tools and task-specific knowledge-acquisition methods. In paper, we describe Web-based browsing system with following features: (a) well-organized English-like presentation concept descriptions (b) use graphs enter relationships, add/delete lists, analogical correspondences. No...